Zhihu and Mianbi Technology jointly released the Chinese model "Zhi Hai Chart AI": the performance of certain scenarios is the same as that of GPT-4, and internal testing begins today.

Thank you, Mr. Air, a netizen of CTOnews.com, for your clue delivery! CTOnews.com April 13 news, today at Zhihu Discovery Conference, Zhihu founder and CEO Zhou Yuan announced that Zhihu and Mianbi technology cooperation of the large Chinese model "Zhi Hai Chu AI" officially started internal testing. At the same time, the "Hot list Summary" based on artificial intelligence starts internal tests to capture, organize and aggregate the questions and answers on the Zhihu hot list, and show the outline of the answers to the users.

Panoramic View of Wall-facing ability according to the introduction of Wall-facing Technology, the large model of "AI" introduced by Zhihu Discovery Conference today is completed by the joint training of wall-facing intelligence and Zhihu. The training of "Zhi Hai Chart AI" is based on the CPM enterprise large model and ModelForce large model system independently developed by face wall intelligence. Based on the "AI" big model, both sides try to explore how to apply the ability of the big model to the Zhihu hot list. For all the answers to Zhihu's hot list of questions, "AI" can quickly extract elements, sort out opinions and aggregate content, and finally show the outline of all answers to Zhihu users, so that everyone can have a faster and more comprehensive understanding of the hot topics and focuses discussed by their friends. In this particular scenario, the effect of the "knowledge chart AI" large model is compared with that of GPT-4, and the two models are almost the same.

CTOnews.com noted that Zhihu US stocks were up 15 per cent in pre-market trading on the news.